According to Bloomberg, the AI lab Z.ai, known for developing the GLM series of models, is the entity behind the recently launched Ox Alpha model. Ox Alpha was anonymously introduced on OpenRouter, where it quickly gained attention for outperforming leading models on various benchmarks and leaderboards. Z.ai has confirmed that Ox Alpha represents the latest iteration of its GLM series, which was previously utilized by Hugging Face to counter an attack from OpenAI agents.
Z.ai plans to publicly release the weights for Ox Alpha on an upcoming Wednesday, enabling developers to build upon the model. The company characterizes Ox Alpha as a reasoning model tailored for coding, sustained agentic work, and production workloads, emphasizing its suitability for long-horizon software engineering, complex reasoning, and workflows integrating text and visual context. The emergence of Ox Alpha contributes to the growing trend of cost-effective, high-capacity models originating from China, potentially challenging the market dominance of expensive frontier model providers such as OpenAI and Anthropic. This development follows Z.ai’s recent release of GLM-5.3, which has demonstrated competitive performance against Anthropic’s Fable 5 on select benchmarks.